This paper deals with the queuing system M/M/1 with additional servers for a longer queue. Clearly the traffic intensity for this system will depend on the number of additional servers. The expected number of customers in the system, the probability of the additional of one server and the probability of the additional of two servers are obtained under the assumption that the number of additional servers depends on the number of customers in the system. The condition under which the M/M/1 queuing system with additional servers is profitable is discussed. A MATLAB program is used to illustrate this condition numerically. Finally, the maximum likelihood estimators of the parameters for this queuing system are obtained.

1. | IJMER | ISSN: 2249–6645 | www.ijmer.com | Vol. 4 | Iss. 5| May. 2014 | 1 |
The Queue M/M/1 with Additional Servers for a Longer Queue
G. S. Mokaddis1
, S. A. Metwally2
, B. E. Farahat3
1, 2, 3,
(Ain Shams University, Department Of Mathematics, Faculty of Science, Cairo, Egypt)
I. Introduction
In many queuing situations, normally the number of servers is dependent on the queue length. For
instance, in a bank more and more windows are opened for service when the queues in front of the already open
windows get too long. This procedure is sometimes used even in the case of airlines, buses, etc. The basic
advantage in such a system is the inherent flexibility of service. Here we shall consider a simplified system of
this type.
Suppose we deal with a queuing system in which the arrivals are individually in a Poisson process with
parameter , and service times of customers are negative exponential with mean µ-1
. The first customer to come
is the first to be served. As long as there are enough customers for service. As long as the number of customers
in the system is greater than or equal to zero and less than or equal to N, there is only one server in the system.
As the number of customers in the system increases to more than N and is still less than or equal to 2N, an
additional server is added. This additional server is removed when the number of customers in the system
decreases to N or less. As soon as the number of customers in the system goes beyond 2N, the number of servers
will be three. Similarly, the third server will be taken off when the number of customers falls to 2N or below.
Let n and µn be the arrival and service rates, respectively, when there are n customers in the system and so, for
the described queuing model, we have
n = for every n≥0 ,
µn=
where, , µ and be the arrival rate, service rate and traffic intensity respectively.
II. System Equations
Assuming the limiting distribution of the number of customers in the system to be { }. We obtain the
following steady-state equations.
 ,
( ) , 1 n N ,
( ) ,
( ) , N n 2N ,
( ) ,
( ) , n 2N .
Abstract: This paper deals with the queuing system M/M/1 with additional servers for a longer queue.
Clearly the traffic intensity for this system will depend on the number of additional servers. The expected
number of customers in the system, the probability of the additional of one server and the probability of
the additional of two servers are obtained under the assumption that the number of additional servers
depends on the number of customers in the system. The condition under which the M/M/1 queuing system
with additional servers is profitable is discussed. A MATLAB program is used to illustrate this condition
numerically. Finally, the maximum likelihood estimators of the parameters for this queuing system are
obtained.
Keywords: Queuing system, traffic intensity, additional servers, customers, maximum likelihood
estimate, queuing length.

2. The Queue M/M/1 with Additional Servers for a Longer Queue
| IJMER | ISSN: 2249–6645 | www.ijmer.com | Vol. 4 | Iss. 5| May. 2014 | 2 |
Solution of the System Equations
This system of equations can be solved recursively, or by using the general solution for single
Markovian queuing models which is given as follows
For 0<n≤N ;
=
= (/µ) n
: (writing /µ = )
= . (2.1)
For N<n≤2N ;
=
= [ ]
= (/µ)N
.(/2µ)n-N
= . (2.2)
For 2N<n ;
=
=
[
]
= (/µ)N
.(/2µ)N
.(/3µ)n-2N
= . (2.3)
Now, to determine we can use the normalizing condition and equations (2.1),(2.2) and
(2.3) as follows
1= + + = + +
= + +
= + +
= ,
= . (2.4)
III. Profitable of using Additional Servers
Suppose the cost to the system due to the presence of a customer is $C1 and the cost of bringing in an
additional server is $C2. It is profitable to use the additional server only if its cost is less than that due to the
waiting customers see [4].
Let Q be the long-run queue length of the standard queue M/M/1 with the same parameters. It is evident that the
M/M/1 queuing system with additional servers is recommended only if
C1[E(Q)-E(Q1)]>C2[Pr(N<Q1≤2N)+2Pr(2N<Q1)], (3.1)
where,
E(Q)=Expected number of customers in the system for the queuing model M/M/1 with no additional
servers (standard).
E(Q1)=Expected number of customers in the system for the queuing model M/M/1 with additional servers.
We have

3. The Queue M/M/1 with Additional Servers for a Longer Queue
| IJMER | ISSN: 2249–6645 | www.ijmer.com | Vol. 4 | Iss. 5| May. 2014 | 3 |
E(Q) =
=
= . (3.2)
Also,
E(Q1)= = + + .
(3.3)
Now we have to find
=
=
= [1+N -(N+1) ].
(3.4)
Similarly,
=2N
[ ]=
[2(N+1) – 2-N+1
(2N+1) – N + 2-N+1
N ]. (3.5)
And
= [ ]=
[3(2N+1) – 2N .
(3.6) Substituting by (3.4) , (3.5) and (3.6) into (3.3) it follows that
E(Q1) = { [1+N -(N+1) ]
+ [2(N+1) – 2-N+1
(2N+1) – N + 2-N+1
N ]
+ [3(2N+1) – 2N }
= [ ,
(3.7)
where
A= 2N
(2- )2
(3- )2
[1+N -(N+1) ]
+ 2N
(1- )2
(3- )2
[2(N+1) – 2-N+1
(2N+1) – N + 2-N+1
N ]
+ (1- )2
(2- )2
[3(2N+1) – 2N ] .
Also, we have to find the probability of the additional of one server,
Pr ( ) = =

4. The Queue M/M/1 with Additional Servers for a Longer Queue
| IJMER | ISSN: 2249–6645 | www.ijmer.com | Vol. 4 | Iss. 5| May. 2014 | 4 |
=
= . (3.8)
And the probability of the additional of two servers ,
Pr ( ) =
=
=
= . (3.9)
Substituting by (3.2), (3.7), (3.8) and (3.9) into (3.1) it follows that the modified M/M/1 queuing system is
recommended only when
<
<
<
< .
Substituting by (2.4), we get
< [ ]
[ ]
< [ ]
[ ]
< [ ].
Substituting by A from equation (3.7)
< {
- (3(2N+1) – 2N
- 2N
(1- )2
(3- )2
[2(N+1) – 2-N+1
(2N+1) – N +2-N+1
N ]} /
.
This can be simplified to

5. The Queue M/M/1 with Additional Servers for a Longer Queue
| IJMER | ISSN: 2249–6645 | www.ijmer.com | Vol. 4 | Iss. 5| May. 2014 | 5 |
<
{
}. (3.10)
From this inequality, depending on the decision variable, the largest value of C2/C1 can be determined when
and N are given. Table (1) has been obtained from the inequality (3.10) for the purpose of illustration.
MATLAB program is used to perform these calculations.
Table (1)
0.50.40.30.20.1/N
6.97785.92935.17624.61194.17651
12.080710.05808.58727.46206.56712
16.556213.663911.585210.02058.80323
20.723317.073614.471812.528611.02644
24.776920.426517.334615.029613.24865
From this table, we can conclude that for any given value of N, the largest value of / decreases
with decreasing values of , that is with increasing values of service rates. The basic advantage in such a
system is the inherent flexibility of service. Then there exists an optimal stationary policy (resulting in the least
long-run expected cost) characterized by a single positive finite number N such that it is optimal to use the slow
service rate when the number of customers in the system is less than or equal to N, faster service rate
when the number in the system is greater than N or less than or equal to 2N and more fast service rate when
the number in the system is greater than 2N.
III. Maximum Likelihood Estimators of The Parameters
To find the likelihood function it is necessary to find the three basic components see [5].
1. the stationary distribution of number of units in the system with contribution ,
2. the interarrival times of length ti [each of which is exponential for n arrival units with contribution
( )] ,
3. the service time of durations ti for k units with contribution ( ) .
Hence, the likelihood function may be written as.
= ; ,
= ;
(use equation (2.3))
= ,
where ,
T= , t= and k1 is a constant .
Thus
=
By partial differentiation with respect to and , respectively, and equating the results by zeros, one can
obtain

6. The Queue M/M/1 with Additional Servers for a Longer Queue
| IJMER | ISSN: 2249–6645 | www.ijmer.com | Vol. 4 | Iss. 5| May. 2014 | 6 |
= =0 (4.1)
and
= =0 , (4.2)
where and QUOTE are the maximum likelihood estimators for the parameters and , respectively.
To obtain and we need to use the following important theorem
THEOREM (4.1),
For the birth-death process with rates, namely 0 , for n positive integer and
both are bounded functions of n [1],
We have
= ,
where = delay probability and
E(n)= expected value of n .
Proof: It is clear that
=
=
Case (i): putting , it follows that
= .
Taking the logarithm of both sides of this equation and differentiating the resulting equation partially with
respect to
= .
But,
= .
Therefore, it follows that
= .
Case (ii): putting , it follows that
= .
Taking the logarithm of both sides of this equation, it follows that
= .
Therefore,
= .
Hence,
= .
Thus, it is evident that
= ,
Hence, the theorem is proved.
Using theorem (4.1), equation (4.1) can be written in the form
,
. (4.3)
Also, by using theorem (4.1), equation (4.2) can be written in the form
,
,

7. The Queue M/M/1 with Additional Servers for a Longer Queue
| IJMER | ISSN: 2249–6645 | www.ijmer.com | Vol. 4 | Iss. 5| May. 2014 | 7 |
. (4.4)
Where is the maximum likelihood estimator of E(n) .
Obtaining the value of from each of the equations in (4.3), (4.4) and dividing the resulting values yields
From (4.3)
From (4.4)
,
,
. (4.5)
Recalling that , the maximum likelihood estimator of the traffic intensity , is given by and
using equations (4.3), (4.4)
(4.6)
This equation can be written in the form
. (4.7)
Consider the special case M/M/1 on replacing by into equation (4.7)
,
,
0 . (4.8)
Replacing (k-u-1) and (n+u+1) by (k-u) and (n+u), respectively, in (4.8) yields.
,
, (4.9)
.
Then
(4.10)
is a solution for the modified equation (4.9) of degree two in . Using (4.5) and (4.10) it follows that
,
(4.11)
To find , put into equation (4.11)
(use equation(4.10))
. (from equation(4.10))
Thus

8. The Queue M/M/1 with Additional Servers for a Longer Queue
| IJMER | ISSN: 2249–6645 | www.ijmer.com | Vol. 4 | Iss. 5| May. 2014 | 8 |
. (4.12)
To find put into equation (4.12)
,
. (4.13)
Then, by using (4.10) it follows that
,
(4.14)
.
REFERENCES
[1]. Feller W. An Introduction to Probability Theory and its Applications ,Vol. John Wiley, New Yourk,1972.
[2]. Gross, D., and Harris, C.M. Fundamentals Of Queuing Theory. John Wiley, New York, 1974.
[3]. Leonard, K. Queuing System, vol. I. John Wiley, New York , 1975 .
[4] Mokaddis , G. S and Zaki. S. S., The Problem Of The Queuing System M/M/1 With Additional Servers For A
Longer Queue. Indian J . Pure Applied Math.1 Vol. 3, No. 3, PP.78-98, 2000.
[5]. Mokaddis, G.S. “ The Statistical Point Estimation for Parameters of Some Queuing Systems” The Sixth International
Congress for Statistical ,Computer Science ,Egypt,29 March-2 April pp.71-93,1981.
[6]. Narayan, U.B. An Introduction to Queuing Theory ,Boston Basel, Berlin 2008.